JP5658588B2 - Hearing presence evaluation device and hearing presence evaluation program - Google Patents

Description

  The present invention relates to an auditory realistic sensation evaluation apparatus and an auditory realistic sensation evaluation program for evaluating auditory realistic sensations based on acoustic signals.

As a technique useful for preprocessing of speech recognition when there are a plurality of speakers, autonomous control of a robot, and the like, research on estimation of a sound source direction has been performed (see Non-Patent Document 1).
Such a sound source direction estimation technique performs microphone array signal processing including multi-channel microphones, and requires a large-scale device to arrange a plurality of microphones.

On the other hand, humans have achieved a sufficiently accurate estimation of the sound source direction using two-channel acoustic signals that can be heard by the left and right ears, and simulated the sound source direction using two-channel acoustic signals. Research to be performed has also been conducted (see Non-Patent Document 2).
Although the sound source direction estimation technique using such two-channel acoustic signals can reduce the size of the apparatus, since the amount of information is small, complicated signal processing is required to ensure estimation accuracy. (See Non-Patent Document 3).

In addition, the conventional sound source direction estimation technique using two-channel acoustic signals can only estimate the direction of one sound source, and the estimation target is limited to a stationary sound source (see Patent Document 1). .
That is, the conventional sound source direction estimation technique tends to be large-scale as a device because it aims at high accuracy, and real-time processing is difficult, and only one sound source direction can be estimated. Had.

On the other hand, PEAQ (Perceptual Evaluation of Audio Quality) has been developed as an objective evaluation method for encoded sound, and this evaluation method has been standardized by ITU (International Telecommunications Union). (ITU-R BS. 1387).
Although this evaluation method imitates the human auditory peripheral system, a simple neural network is substituted for the part of the auditory central system.

  In addition, as an objective evaluation method for voice quality perceived by humans, there is an evaluation method described in Patent Document 2. Such an evaluation method is a method for objectively evaluating the degree of deterioration related to a reference sound.

Japanese Patent Laid-Open No. 5-87903 JP 2004-172753 A

Arthur N. Popper, Richard R. Fay, Eds., "Sound Source Localization", Springer Handbook of Auditory Research, 2005, Springer, New York Hitoshi Nagata, “Computation Model for Estimating Direction of Arrival Sound Based on Two-Channel Signals”, Comparative Physiological and Biochemical Society of Japan, 2010, Vol. 27, No. 1, p.10-18 Nicoleta Roman, DeLiang Wang, “Binaural Tracking of Multiple Moving Sources”, IEEE Transactions on Audio, Speech, and Language Processing, 2008, Vol. 16, No. 4, p. 728-739

  Along with the development of sound collection and playback technology, many acoustic systems such as 22.2 multi-channel acoustic system and Wave Field Synthesis have been developed that can realize a high level of realism, improving the quality, not the degree of sound quality degradation. An objective evaluation of the degree is required.

  Here, K. Ozawa, Y. Chujo, Y. Suzuki, T. Sone, "Contents which yield high auditory-presence in sound reproduction", Kansei Engineering International, 2002, Vol. 3, No. 4, p. 25 -30, and K. Ozawa, Y. Chujo, “Content Presence vs. System Presence in Audio Reproduction Systems”, Proc. Of the Second International Symposium on Universal Communication (ISUC2008), 2008, p. 50-55 As described above, it has been clarified that the listener feels that there is a sense of reality as the sound image of the acoustic signal reproduced by the acoustic device moves relative to the listener.

  As a result of diligent research, the inventor of the present application may objectively evaluate the degree of improvement in sound quality by quantifying the auditory presence, that is, the auditory presence. I came up with the idea.

  The present invention has been made in view of the above-described circumstances, and an object thereof is to provide an auditory realistic sensation evaluation apparatus and an auditory realistic sensation evaluation program capable of objectively evaluating auditory realistic sensations.

  In order to solve the above problems, an auditory presence evaluation apparatus according to the present invention is an auditory presence evaluation apparatus that evaluates an auditory presence based on two acoustic signals measured by two microphones, and includes an acoustic signal dividing unit. And a cross-correlation function calculation unit, a moving image generation unit, a sound image movement vector calculation unit, and an auditory realistic sense value calculation unit.

According to such a configuration, the two acoustic signals measured by the acoustic signal dividing unit are cut out by, for example, a certain section length every certain time, divided for each frequency band, and divided by the cross-correlation function calculating unit. Using the two acoustic signals, a cross-correlation function related to the two acoustic signals is calculated for each frequency band. Then, any of luminance, hue, and saturation of the two-dimensional moving image with the cross-correlation function calculated by the moving image generation unit as the coordinate axis of the frame with the lag of the two acoustic signals and the frequency band as the frame coordinate axes. generating a moving image by crab conversion, calculated by the sound image movement vector calculating unit, by calculating a movement vector of the moving image using a plurality of frames of the generated the moving image, the moving vector of the sound image Then, based on the calculated sound image movement vector, the auditory presence evaluation value calculation unit calculates the auditory presence feeling evaluation value such that the larger the sound image movement vector is, the larger the auditory presence feeling evaluation value is. be able to.

  The moving image generation unit may be configured to convert the cross-correlation function into the luminance. According to this configuration, it is possible to generate a monochrome moving image having a luminance component that correlates with the cross-correlation function, and to calculate an auditory presence sense evaluation value from the movement vector of the generated monochrome moving image.

  The auditory presence evaluation apparatus may further include a sound pressure level calculation unit.

  According to such a configuration, the sound pressure level calculation unit calculates the sound pressure levels of the two acoustic signals for each frequency band using the divided two acoustic signals, and is calculated by the moving image generation unit. The moving image can be generated by converting the sound pressure level into any one of the luminance, the hue, and the saturation other than the one obtained by converting the cross-correlation function. In other words, the auditory realistic sensation evaluation apparatus calculates the auditory realistic sensation evaluation value using more parameters, so that the calculation accuracy of the auditory realistic sensation is improved.

  The moving image generation unit may be configured to convert the cross-correlation function into the luminance and convert the sound pressure level into the saturation. According to this configuration, a color moving image having a luminance component that correlates with a cross-correlation function and a saturation component that correlates with a sound pressure level is generated, and an auditory presence evaluation value is calculated from a movement vector of the generated color moving image. can do.

  Further, the sound pressure level calculation unit calculates a sound pressure level difference, which is a difference between the sound pressure levels, for each frequency band, and the moving image generation unit calculates the calculated sound pressure level difference as the luminance. The moving image may be generated by converting the hue and the saturation other than those obtained by converting the cross-correlation function and the sound pressure. In other words, the auditory realistic sensation evaluation apparatus calculates the auditory realistic sensation evaluation value using more parameters, so that the calculation accuracy of the auditory realistic sensation is improved.

  Furthermore, the moving image generation unit may be configured to convert the cross-correlation function into the luminance, convert the sound pressure level into the saturation, and convert the sound pressure level difference into the hue. . According to such a configuration, a color moving image having a luminance component correlated with the cross-correlation function, a saturation component correlated with the sound pressure level, and a hue component correlated with the sound pressure level difference is generated, and the generated color moving image An auditory realistic sense evaluation value can be calculated from the movement vector.

  The cross-correlation function calculating unit may be configured to calculate a plurality of cross-correlation functions by section length having different section lengths, and calculate the cross-correlation function based on the calculated cross-correlation functions by section length. Good. When only the cross-correlation function having a single section length is calculated, the value of the cross-correlation function that is calculated by chance is high and the actual sound image position may not exactly match. By using a plurality of section lengths, erroneous detection of the sound image position can be prevented.

  The sound image movement vector calculation unit may be configured to calculate a movement vector of the moving image using three or more frames of the moving image. According to this configuration, the calculation accuracy of the movement vector is improved.

  The present invention can also be embodied as an auditory reality evaluation program that causes a computer to function as the aforementioned auditory reality evaluation device.

  According to the present invention, auditory presence can be objectively evaluated.

It is a block diagram which shows the auditory presence sense evaluation system which concerns on embodiment of this invention. It is a graph which shows the acoustic signal measured by a pair of microphone. (A) is a graph which shows a cross correlation function, (b) is a figure which shows the relationship between the value of a cross correlation function, and a brightness | luminance, (c) shows the pixel row | line | column produced | generated from the cross correlation function. FIG. (A) is a graph which shows a sound pressure level difference, (b) is a graph which shows the weight by sound pressure level difference, (c) is a figure which shows the weight with respect to a pixel row | line | column. (A) is a figure which shows a moving image, (b) is a figure which shows a movement vector. (A)-(d) is a figure for demonstrating the calculation method of sound image moving amount | distance. (A) (b) is a figure for demonstrating the example of calculation of the movement vector of a single sound image. (A) (b) is a figure for demonstrating the example of calculation of the movement vector of a several sound image.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings as appropriate, taking as an example the case where the present invention is applied to evaluation of a reproduction sound field. Similar parts are denoted by the same reference numerals, and redundant description is omitted. In the present invention, the “sound image” refers to a position where a listener perceives the presence of a sound source, and may be a virtual sound source position embodied by a stereo speaker or the like, or an actual sound source position. In the present invention, “auditory presence” refers to the sense of presence that a listener feels in hearing. Conventionally, the sense of presence has been used in an ambiguous sense even though it is an important keyword when talking about the performance of an AV device. The value can be estimated quantitatively and objectively.

  The playback device includes a stereo speaker, a 5.1 channel system, headphones, and the like, but the listener perceives a sense of auditory presence with an acoustic signal that finally reaches both ears. Therefore, the auditory realistic sensation evaluation apparatus of the present invention estimates the moving amount of the sound image based on the two-channel acoustic signal, similarly to the listener's ears.

  As shown in FIG. 1, the auditory presence evaluation system 1 according to the embodiment of the present invention includes a storage medium playback device 10 and a speaker group 20 as playback-side devices. The storage medium playback device 10 reads data stored in a storage medium (not shown) and plays back an acoustic signal via a speaker group 20 including a stereo speaker, a 5.1 channel system, headphones, and the like.

  The auditory presence evaluation system 1 according to the embodiment of the present invention includes a pair of microphones 30L and 30R, an auditory presence evaluation device 40, and a notification device 50 as devices on the evaluation side.

<Microphone>
The pair of microphones 30 </ b> L and 30 </ b> R has a two-channel structure arranged on the left and right sides, measures an acoustic signal reproduced by the speaker group 20, and outputs the measured two-channel acoustic signal to the auditory presence evaluation device 40. . In this specification, not only the sound waves reproduced by the speaker group 20 but also the measurement results of the sound waves by the microphones 30L and 30R are described as acoustic signals.

<Hearing presence evaluation device>
The auditory presence evaluation device 40 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read-Only Memory), an input / output circuit, and the like. , A cross-correlation function calculating unit 42, a sound pressure level calculating unit 43, a moving image generating unit 44, a sound image movement vector calculating unit 45, and an auditory realistic sense value calculating unit 46.

≪Sound signal division part≫
The acoustic signal dividing unit 41 acquires the two-channel acoustic signals output from the pair of microphones 30L and 30R for a certain section length at certain time intervals, and the acquired two-channel acoustic signals are M bandpass filters. Is divided into M for each frequency band, and the divided acoustic signals are output to the cross-correlation function calculator 42 and the sound pressure level calculator 43.

In the present embodiment, the acoustic signal dividing unit 41 includes five octave bandpass filters, and the acoustic signal L output from the microphone 30L is converted into the _first frequency band f1 having a center frequency of 125 [Hz]. One acoustic signal L ₁ , a second acoustic signal L ₂ in a frequency band f ₂ with a center frequency of 250 [Hz], and a third acoustic signal L ₃ in a frequency band f ₃ with a center frequency of 500 [Hz]. The fourth acoustic signal L ₄ in the frequency band f ₄ with a center frequency of 1000 [Hz] is divided into the fifth acoustic signal L ₅ in the frequency band f ₅ with a center frequency of 2000 [Hz].

Similarly, the acoustic signal division unit 41 converts the acoustic signal output from the microphone 30R into a first acoustic signal R ₁ in the frequency band f ₁ having a center frequency of 125 [Hz] and a center frequency of 250 [Hz]. A second acoustic signal R _{2 in} a certain frequency band f _2, a third acoustic signal R ₃ in a frequency band f ₃ having a center frequency of 500 [Hz], and a frequency band f ₄ having a center frequency of 1000 [Hz]. The fourth acoustic signal R ₄ is divided into a fifth acoustic signal R ₅ in a frequency band f ₅ having a center frequency of 2000 [Hz].

  Note that the number of divisions of the acoustic signal is not limited to five, and can be appropriately changed according to the type of the target sound image. For example, when targeting a sound including a broadband noise component such as automobile noise, M = 10 is obtained by dividing an audible frequency band (20 [Hz] to 20 [kHz]) by a 1/1 octave bandwidth. It can be. When a sound having a harmonic structure such as a musical instrument sound is targeted, M = 40 divided by 1/10 octave bandwidth can be set. As a general-purpose model band-pass filter, an auditory filter of about 50 channels (a band-pass filter that takes into account the characteristics of the auditory peripheral system, representative examples include a ROEX filter, a Gamma-tone filter, etc.) By adopting, the division reflecting the auditory characteristics can be performed.

≪Cross-correlation function calculation part≫
The cross-correlation function calculating unit 42 acquires the acoustic signals L _{1 to} L ₅ and R _{1 to} R ₅ output from the acoustic signal dividing unit 41, and acquires the acquired acoustic signals L _{1 to} L ₅ , R ₁ to Based on R ₅ , a cross-correlation function for each frequency band is calculated, and the calculated cross-correlation function is output to the moving image generation unit 44.

In the present embodiment, the cross-correlation function calculation unit 42 calculates a cross-correlation (CC) function of the acoustic signals L ₁ and R ₁ for each calculation period (time t ₁ , t ₂ ,...).
As shown in FIG. 2A, the cross-correlation function calculating unit 42 extracts and extracts the acoustic signals L ₁ and R ₁ having a section length (also referred to as a time window width) T ₁ divided from the time t ₁ as the head. 3 (a) by calculating the correlation value of the acoustic signals L ₁ and R ₁ and repeating the calculation of the correlation value by shifting and extracting only the acoustic signal R ₁ by shifting it by lag m. As shown in the lower stage, a cross-correlation function with lag m as a variable is calculated. The cross-correlation function takes a value from −1 to +1 and becomes +1 when the acoustic signals L ₁ and R ₁ coincide with each other, and the value decreases as the waveforms of the acoustic signals L ₁ and R ₁ differ. In the present invention, the lag m is a section shift (time difference) for calculating a cross-correlation function in the acoustic signals L ₁ and R ₁ .

An acoustic signal L _{p, q} (n), R _p that is a function of time n in the section length T ₁ , p-th calculation period, and q-th frequency band divided by the above-described times t ₁ , t ₂ ,. _{, Q} (n), the cross-correlation function r _{p, q} (m) is calculated by the following equation.

Here, σ ^L _{p, q} and σ ^R _{p, q} are standard deviations of L _{p, q} (n) and R _{p, q} (n), respectively. N is a discrete time corresponding to the length T of the period for calculating the cross-correlation function.

  The section length can be changed as appropriate according to the type of target sound image. For example, when the target is stationary noise, it is set to about 100 to 200 [ms], and when the target is instrumental sound having a large temporal change, it is set to about 10 to 20 [ms]. it can. The section length may be stored in advance in the cross-correlation function calculation unit 42, and the user operates an input device (not shown) including a keyboard and a mouse connected to the auditory presence evaluation device 40. The configuration may be set.

Further, the calculation cycle (interval between times t ₁ , t ₂ ,...) And the section length may or may not match.

Further, the cross-correlation function calculation unit 42 may be configured to change the calculation cycle based on the moving amount of the sound image (the magnitude of the movement vector) calculated by the auditory realistic sense value calculation unit 46. For example, when the moving amount of the sound image is small, the calculation amount can be reduced by setting the calculation cycle longer. In this case, even when the cross-correlation function is calculated using 1/30 [s], which is a general frame rate of a two-dimensional moving image, as a calculation cycle, the section length is longer than 1/30 [s]. set, it is possible to calculate the cross-correlation function while overlapping extraction range of the acoustic signal R _1. The cross-correlation function calculation unit 42 stores in advance a relationship between the amount of movement of the sound image and the calculation cycle of the cross-correlation function as a database. By searching the database using the movement amount of the sound image calculated by the above, a calculation cycle corresponding to the movement amount of the sound image is read, and a cross-correlation function is calculated based on the read calculation cycle.

The cross-correlation function calculating unit 42 is configured to calculate a plurality of cross-correlation functions by section length having different section lengths, and to calculate a cross-correlation function based on the calculated plurality of cross-correlation functions by section length. Also good. For example, the cross-correlation function calculation unit 42 stores section lengths T ₁ , T ₂ , and T ₃ in advance, and the cross-correlation function calculation unit 42 stores the section length T _{1 as} shown in FIG. , The cross-correlation function by section length in section lengths T ₂ and T ₃ longer than the section length T ₁ is calculated, and the calculated three cross-correlation functions by section length are calculated. By calculating the arithmetic mean, the cross-correlation function can be calculated.

The cross-correlation function calculating unit 42 similarly calculates the cross-correlation function for the other frequency bands f _{2 to} f ₅ .

≪Sound pressure level calculation part≫
The sound pressure level calculation unit 43 acquires the acoustic signals L _{1 to} L ₅ and R _{1 to} R ₅ output from the acoustic signal division unit 41, and acquires the acquired acoustic signals L _{1 to} L ₅ , R ₁ to Based on R ₅ , the sound pressure level for each frequency band and the sound pressure level difference that is the difference between the sound pressures of the left and right acoustic signals are calculated, and the calculated sound pressure level and the sound pressure level difference are generated as a moving image. And outputs the calculated sound pressure level to the auditory presence sense value calculation unit 46.

≪Moving image generation part≫
Based on the cross-correlation function for each frequency band output from the cross-correlation function calculation unit 42 and the sound pressure level and the sound pressure level difference output from the sound pressure level calculation unit 43, the moving image generation unit 44 (More specifically, a two-dimensional moving image) is generated, and the generated moving image is output to the sound image movement vector calculation unit 45.

As shown in FIG. 5 (a), moving images, the frame 100 at time _{t 1,} the frame 200 at time _{t 2, the} frame 300 at time _{t 3,} the frame 400 at time _{t 4,} ... are composed of, videos One frame of the image is a plane in which the vertical axis represents frequency (frequency band f _{1 to} f ₅ ) and the horizontal axis represents lag m. In the present embodiment, one frame is composed of 5 × 8 pixels, and the moving image generation unit 44 uses the cross-correlation function for the luminance of the corresponding pixel and the saturation of the pixel corresponding to the sound pressure level. The moving image is generated by converting the sound pressure level difference into the hue of the corresponding pixel.

For example, when a moving image is generated using only the cross-correlation function, the moving image generation unit 44 has the lowest luminance when the cross-correlation function is −1 and the highest luminance when the cross-correlation function is +1. In this way, pixel data of each pixel is generated (see FIG. 3B), and a moving image composed of a series of images constituted by the pixel data is generated. Here, moving image generation unit 44, the cross-correlation function of the frequency band f ₁ shown in the lower part of FIG. 3 (a), to generate a pixel column 110 having a luminance as shown in the lower part of FIG. 3 (c), FIG. 3 from the cross-correlation function of the frequency band f ₂ shown in the upper part of (a), to generate a pixel column 120 having a luminance as shown in the upper part of FIG. 3 (c).

In addition, the moving image generation unit 44 has the sound pressure levels 210 and 220 of the acoustic signals L ₁ (or L _{2 to} L ₅ ) and R ₁ (or R _{2 to} R ₅ ) for each frequency band shown in FIG. Based on the sound pressure level difference 230, a weight for each lag m shown in FIG. 4B may be calculated, and a pixel row 110 ′ having luminance weighted by the calculated weight may be generated. .

  When a highly correlated acoustic signal arrives with a sound pressure level difference between the left and right ears, a sound image is perceived in any direction of ± 90 degrees with the front as 0 degrees due to the left and right level differences. The moving image generation unit 44 stores the relationship between the sound pressure level difference and the perceived direction of the sound image in advance, and refers to the relationship using the sound pressure level difference 230, whereby the sound image is generated in any number of directions. You can guess if there is. The moving image generating unit 44 sets 1 as a weight in a direction in which it is estimated that there is a sound image, and sets a small value as a weight in a direction away from the moving image generation unit 44.

On the other hand, since the lag m that maximizes the value of the cross-correlation function is the direction in which the sound image exists, the lag m and the direction of the sound image correspond one-to-one. Therefore, the moving image generation unit 44 keeps the value of the cross-correlation function in the vicinity estimated that there is a sound image based on the sound pressure level difference 230 by weighting, and if there is no sound image based on the sound pressure level difference 230. The value of the estimated cross-correlation function in the vicinity is made small. By doing so, the moving image generation unit 44 determines the direction of the sound image based on the sound pressure level difference 230 even when there are several lags where the cross-correlation function is 1, such as a sine wave. can do.

  Further, the moving image generation unit 44 is configured to generate a moving image by regarding a portion having a negative cross correlation function as 0 when generating a moving image by converting the cross correlation function into luminance or hue. Also good. This is because when there is one sound image, the position corresponding to the lag at which the cross-correlation function is maximized matches the position of the sound image.

≪Sound image movement vector calculation unit≫
The sound image movement vector calculation unit 45 calculates a movement vector of the sound image by calculating a movement vector of the moving image using a plurality of frames of the moving image output from the moving image generation unit 44, and calculates the sound image of the calculated sound image. The movement vector is output to the auditory presence evaluation value calculation unit 46. As a method for calculating a moving vector of a moving image, a known algorithm for image information processing such as optical flow, pattern matching, feature amount tracking algorithm, or the like can be appropriately employed.

The sound image movement vector calculation unit 45 accumulates three or more (in this embodiment, _four frames corresponding to times t _{1 to} t 4), and based on the luminance, hue, and saturation of the accumulated frames, A movement vector for each frequency band is calculated. Specifically, as shown in FIG. 5B, the sound image movement vector calculation unit 45 has a pixel column 111 at time t ₁ (here, the number of pixels in one column is 12 for explanation), pixel row 211 at time _{t 2, the} using the pixel column 311 at time _{t 3,} and calculates a movement vector in the frequency band _{f 1,} the same processing is performed for the other frequency band _f 2 ~f _5. Here, when the movement vector is calculated using the two frames at times t ₁ and t ₂ , the sound image movement vector calculation unit 45 determines that the fourth pixel from the left of the pixel column 111 at the time t ₁ , there is a risk of an erroneous calculation remains the 4 th left pixel column 211 at time t ₂ (solid arrow). Therefore, the sound image movement vector calculation unit 45 further uses the pixel column 311 at the time t ₃ , so that the _first to third pixels from the left of the pixel column 111 at the time t ₁ are in the pixel column 211 at the time t ₂ . 5-7 th from the left, since the pixel rows at a time t ₃ is estimated to be moving to 10 to 12 from the left, 4 th left pixel of the pixel row 111 at time t ₁ is, it can be calculated to be moving from the left pixel column 211 to 8 th at time t ₂ (dashed arrows). Thus, by calculating the movement vector using three or more frames, it is possible to prevent erroneous calculation of the movement vector.

  In addition, the sound image movement vector calculation unit 45 may calculate the movement vector of the sound image so as to straddle the frequency band according to the change in the sound pressure level for each frequency band with respect to the sound image whose frequency characteristics vary. Here, the sound image movement vector calculation unit 45 calculates the peak (ridge portion) of the sound pressure level in a three-dimensional space with the frequency, time, and sound pressure level as axes based on the time change of the frequency component of the sound pressure level. It is possible to determine that the frequency characteristics of the sound image have fluctuated and shift the sound image movement vector so as to straddle the frequency band.

≪Hearing presence evaluation value calculation part≫
The auditory presence sense evaluation value calculation unit 46 calculates an auditory presence sense evaluation value based on the sound image movement vector output from the sound image movement vector calculation unit 45 and outputs it to the notification device 50. In the present embodiment, the auditory realistic sensation evaluation value calculation unit 46 calculates the auditory realistic sensation evaluation value by rounding down the movement amount of the sound image less than the threshold value and arithmetically averaging the movement amounts of the plurality of sound images equal to or greater than the threshold value. can do. The auditory presence evaluation value is a value such that the larger the movement vector of the sound image, the larger the auditory presence evaluation value.

  The auditory presence sense value calculation unit 46 estimates a sound image movement amount for each frequency band based on a movement vector for each frequency band of the moving image, that is, a movement vector for the sound image, and moves the sound image for each estimated frequency band. The overall sound image movement amount is estimated by calculating the arithmetic average of the sound image movement amounts using the maximum value of the amount, or by calculating the integrated value of the sound image movement amounts.

In the present embodiment, the auditory realistic sense value calculation unit 46 employs an integrated value of the sound image movement amount. That is, as shown in FIG. 6A, the auditory presence sense value calculation unit 46 moves the movement vectors V _1a , V _1b , V _{2 to} V _{5 of the} frequency bands f _{1 to} f ₅ in the previous frame and the current frame. Is calculated.

Subsequently, as shown in FIG. 6B, the auditory presence sense value calculation unit 46 integrates the movement vectors of the respective frequency bands f _{1 to} f ₅ . Here, the movement vector V ₅ of the movement vector V ₂ and the frequency band f ₅ of the frequency band f ₂ are integrated because of the similar frequency band f ₁ left movement vector V _1a and the frequency band f ₁ of a movement vector V ₄ of the left movement vector V _1b and the frequency band f ₄ of are integrated because of the similar.

  Here, the auditory presence evaluation value calculation unit 46 determines that the two movement vectors have the same direction and the same magnitude (for example, the ratio of the magnitudes of the movement vectors is within a certain range). The configuration may be such that the vectors are determined to be similar, and further, if at least one of the sound pressure level and the sound pressure level difference between the two movement vectors is substantially the same (the ratio is within a certain range), two The configuration may be such that the two movement vectors are determined to be similar.

Subsequently, as shown in FIG. 6C, the auditory presence sense evaluation value calculation unit 46 sums the magnitudes of the movement vectors V _1b (= V _1a = V ₄ ), V ₃ , V ₅ after integration ( Integral value) is calculated, and the calculation result is used as the moving amount of the sound image.

As shown in FIG. 6D, the auditory presence sense evaluation value calculation unit 46 corrects the integrated movement vectors V _1b , V ₃ , and V ₅ with sound pressures, and corrects the corrected movement vector V _1b ′. , V ₃ ′, V ₅ ′ may be summed (integrated value), and the calculated result may be used as a sound image movement amount. For example, weighting is performed such that the greater the sound pressure level, the higher the contribution degree of the sound image movement vector to the auditory presence evaluation value, thereby improving the estimation accuracy of the auditory presence evaluation value.

  Subsequently, the auditory presence evaluation value calculation unit 46 calculates the sound image movement amount in the current frame and the next frame, repeats the calculation until the predetermined number of frames is reached, and sums the sound image movement amounts for the predetermined number of frames (integrated value). ) And the calculated result is the final sound image movement amount, that is, the auditory presence evaluation value.

In addition, the auditory presence evaluation value calculation unit 46 calculates a plurality of auditory presence evaluation values according to the number of frames having different numbers of frames, and the auditory presence evaluation based on the plurality of calculated auditory presence feeling evaluation values according to the number of frames. The structure which calculates a value may be sufficient.
For example, the auditory reality evaluation value calculation unit 46 calculates the auditory presence evaluation value for each frame number in three frames and the auditory presence evaluation value for each frame number in four frames. By calculating the arithmetic average of the auditory realistic sensation evaluation values according to the number of frames, the auditory realistic sensation evaluation value can be calculated.

  When there is only one sound image, the auditory presence evaluation value calculation unit 46 determines the direction (azimuth angle) of the sound image at a certain time based on the lag m that maximizes the luminance (that is, the cross-correlation function). It can also be estimated. This is because, when there is one sound image, the lag m that maximizes the cross-correlation function is related to the difference in distance from the sound image to each of the microphones 30L and 30R.

  In addition, when the sound pressure levels of a plurality of frequency bands fluctuate synchronously, the auditory presence sense evaluation value calculation unit 46 considers that the same sound image exists in the plurality of frequency bands and determines a movement vector of the sound image. It can also be integrated. According to this configuration, it is possible to calculate an auditory realistic sense evaluation value using a movement vector of a sound image similar to that perceived by a listener.

<Notification device>
The notification device 50 includes a display, a speaker, and the like, and notifies the user by displaying or outputting the auditory reality evaluation value output from the auditory reality evaluation value calculation unit 46.

<Operation example>
Subsequently, an operation example of the auditory presence sense evaluation device 40 according to the embodiment of the present invention will be described. First, the acoustic signal dividing unit 41 obtains two-channel acoustic signals output from the pair of microphones 30L and 30R, and the obtained two-channel acoustic signals have five different frequency bands by five band-pass filters. Are divided into acoustic signals L _{1 to} L ₅ and R _{1 to} R ₅ .

Subsequently, the cross-correlation function calculation unit 42 calculates a cross-correlation function for each frequency band based on the acoustic signals L _{1 to} L ₅ and R _{1 to} R ₅ .

Subsequently, the sound pressure level calculation unit 43 is the difference between the sound pressure level for each frequency band and the sound pressure of the left and right sound signals based on the sound signals L _{1 to} L ₅ and R _{1 to} R _5. The sound pressure level difference is calculated, and the calculated sound pressure level and the sound pressure level difference are output to the moving image generation unit 44, and the calculated sound pressure level is output to the auditory presence evaluation value calculation unit 46.

  Subsequently, based on the cross-correlation function and the sound pressure level and the sound pressure level difference output from the sound pressure level calculation unit 43, the moving image generation unit 44 sets the cross-correlation function to luminance, the sound pressure level to saturation, Pixel data of each pixel, in which the sound pressure level difference is regarded as a hue, is generated, and a moving image including a series of images configured by the pixel data is generated.

  Subsequently, the sound image movement vector calculation unit 45 calculates the movement vector of the sound image by calculating the movement vector of the moving image using the plurality of frames of the moving image output from the moving image generation unit 44.

  Subsequently, the auditory presence evaluation value calculation unit 46 calculates an auditory presence evaluation value based on the sound image movement vector output from the sound image movement vector calculation unit 45, and outputs it to the notification device 50.

<Example of calculation of movement vector of sound image>
Subsequently, an example in which the sound image movement vector calculation unit 45 of the auditory presence evaluation apparatus 40 according to the embodiment of the present invention actually calculates the movement vector of the moving sound image will be described.

  Here, a dummy head (SAMRAI manufactured by Koken Co., Ltd.) is used as the microphones 30L and 30R, and a single speaker is used as the speaker group 20, and the speaker has a radius of about 0. The dummy head detected an acoustic signal when moving from the left (−90 [deg]) to the right (90 [deg]) or from the right to the left on a circle of 5 [m]. The sampling frequency of the acoustic signal dividing unit 41 is 44100 [Hz], and the acoustic signal dividing unit 41 is configured using an octave bandpass filter having center frequencies of 125, 250, 500, 1000, 2000, and 4000 [Hz]. By doing so, the acoustic signal was divided into six frequency bands.

  The cross-correlation function calculation unit 42 calculates the cross-correlation function by setting the cross-correlation function calculation period to 1/30 [s] and the section length to 1/5 [s]. Here, since the distance between the microphones 30L and 30R in the dummy head is 0.18 [m] and the sound speed is 340 [m / s], the time difference between the acoustic signals reaching the microphones 30L and 30R is maximized. The time difference when the speaker is on the left or right is about 0.7 [ms]. Therefore, as the lag m, attention was paid only to a portion corresponding to ± 0.7 [ms] centered on 0.

In addition, the sound image movement vector calculation unit 45 uses a technique of connecting pixels having the maximum values r _{p−1, q} (m) _max , r _{p, q} (m) _max of the cross-correlation function, to a moving image movement vector. That is, the movement vector of the sound image was calculated.

≪Single sound image≫
When white noise is output as an acoustic signal from a single speaker, that is, when a single sound image is generated using a single sound source, an image (frame) generated by the moving image generation unit 44 at the measurement start time is displayed. FIG. 7B shows an image generated by the moving image generating unit 44 at the present time (halfway elapsed time) as shown in FIG. The arrow in FIG. 7B is a moving image of the sound image for each frequency band from the start time to the present. Referring to FIGS. 7A and 7B, it can be seen that the sound image moves from left to right in all frequency bands.

≪In the case of multiple sound images≫
An acoustic signal that outputs low-pass noise (cutoff frequency 700 [Hz]) from a speaker moving from left to right, and an acoustic signal that outputs high-pass noise (cutoff frequency 1000 [Hz]) from a speaker moving from right to left; The synthesized signal is input to the acoustic signal dividing unit 41 for analysis. In this case, two sound images are generated using two sound sources. In this case, an image (frame) generated by the moving image generating unit 44 at the measurement start time is shown in FIG. 8A, and an image generated by the moving image generating unit 44 at the present time (halfway elapsed time) is shown in FIG. Shown in The arrows in FIG. 8B are sound image movement vectors for each frequency band from the start time to the present time. Referring to FIGS. 8A and 8B, it can be seen that the sound image moves from left to right in the low frequency band and the sound image moves from right to left in the high frequency band.

  The auditory presence evaluation device 40 according to the embodiment of the present invention calculates an auditory presence evaluation value based on the moving amount of the sound image without using the direction of the sound image, even when processing a plurality of sound images. Since the calculation is performed, the auditory presence can be objectively evaluated with a simple process.

  As mentioned above, although embodiment of this invention was described with reference to embodiment, this invention is not limited to the said embodiment, A design change is possible suitably in the range which does not deviate from the summary of this invention. For example, without the sound pressure level calculation unit 43, the moving image generation unit 44 may generate a monochrome moving image in which the cross-correlation function is converted to luminance. Alternatively, the moving image generation unit 44 may generate a color moving image having the cross-correlation function as luminance and the sound pressure level as saturation. In this case, the moving image generation unit 44 can set the hue of each pixel to a predetermined value (for example, a hue value of 50%). Alternatively, the moving image may be generated by converting the cross-correlation function, the sound pressure level, and the sound pressure level difference into any one of luminance, saturation, and hue.

  The auditory presence evaluation device 40 can also calculate an auditory presence evaluation value for an apparatus that reproduces an acoustic signal that is not stored in a storage medium or an actual sound source that is not a reproduction apparatus. Further, the sound image movement vector calculation unit 45 may be configured to calculate an integrated movement vector by calculating one movement vector using a pixel array of a plurality of frequency bands. The feature amount may be stored in advance as a pattern, and the movement vector may be calculated using pattern matching. The present invention can also be embodied as an auditory reality evaluation program that causes a computer to function as the auditory reality evaluation device 40.

DESCRIPTION OF SYMBOLS 1 Auditory presence evaluation system 30L, 30R Microphone 40 Auditory presence evaluation apparatus 41 Acoustic signal division part 42 Cross-correlation function calculation part 43 Sound pressure level calculation part 44 Moving image generation part 45 Sound image movement vector calculation part 46 Auditory presence feeling evaluation value Calculation unit

Claims (9)

An auditory realistic sensation evaluation apparatus that evaluates auditory realism based on two acoustic signals measured by two microphones,
An acoustic signal divider that divides the two measured acoustic signals for each frequency band; and
A cross-correlation function calculation unit that calculates a cross-correlation function for the two acoustic signals for each of the frequency bands using the divided two acoustic signals;
A moving image is obtained by converting the calculated cross-correlation function into one of luminance, hue, and saturation of a two-dimensional moving image having the lag of the two acoustic signals and the frequency band as coordinate axes of a frame. A moving image generation unit for generating
A sound image movement vector calculating unit that calculates a movement vector of the sound image by calculating a movement vector of the moving image using a plurality of frames of the generated moving image;
Based on the calculated movement vector of the sound image, the auditory reality evaluation value calculation unit that calculates the auditory presence feeling evaluation value so that the auditory presence feeling evaluation value increases as the movement vector of the sound image increases;
An auditory presence evaluation device comprising: The auditory realistic sensation evaluation apparatus according to claim 1, wherein the moving image generation unit converts the cross-correlation function into the luminance. A sound pressure level calculation unit that calculates the sound pressure level of the two acoustic signals for each of the frequency bands using the divided two acoustic signals,
The moving image generation unit converts the calculated sound pressure level into one of the luminance, the hue, and the saturation other than the one obtained by converting the cross-correlation function. The auditory presence evaluation device according to claim 1, wherein the device is generated. The auditory presence evaluation apparatus according to claim 3, wherein the moving image generation unit converts the cross-correlation function into the luminance and converts the sound pressure level into the saturation. The sound pressure level calculation unit calculates a sound pressure level difference, which is a difference between the sound pressure levels, for each frequency band,
The moving image generation unit converts the calculated sound pressure level difference into a moving image by converting the luminance, the hue, and the saturation other than the one obtained by converting the cross-correlation function and the sound pressure. The auditory presence evaluation apparatus according to claim 3, wherein an image is generated. 6. The moving image generation unit converts the cross-correlation function into the luminance, converts the sound pressure level into the saturation, and converts the sound pressure level difference into the hue. The auditory presence evaluation device according to 1. The cross-correlation function calculating unit calculates a plurality of cross-correlation functions for each section length having different section lengths, and calculates the cross-correlation function based on the plurality of calculated cross-correlation functions for each section length. The auditory presence evaluation device according to any one of claims 1 to 6. The auditory presence according to any one of claims 1 to 7, wherein the sound image movement vector calculation unit calculates a movement vector of the moving image using three or more frames of the moving image. Feeling evaluation device. An auditory presence evaluation program for evaluating auditory presence based on two acoustic signals measured by two microphones,
Computer
An acoustic signal dividing unit that divides the two measured acoustic signals for each frequency band;
A cross-correlation function calculation unit that calculates a cross-correlation function for the two acoustic signals for each of the frequency bands using the divided two acoustic signals,
A moving image is obtained by converting the calculated cross-correlation function into one of luminance, hue, and saturation of a two-dimensional moving image having the lag of the two acoustic signals and the frequency band as coordinate axes of a frame. A moving image generation unit for generating
A sound image movement vector calculating unit for calculating a movement vector of a sound image by calculating a movement vector of the moving image using a plurality of frames of the generated moving image; and
Based on the calculated movement vector of the sound image, an auditory presence evaluation value calculation unit that calculates the auditory presence feeling evaluation value so that the larger the movement vector of the sound image is, the larger the auditory presence feeling evaluation value is;
Auditory presence evaluation program characterized by functioning as